128: LaraProm  🔗 **Source:** [128: LaraProm](https://share.snipd.com/episode/a8488669-f02f-418a-acc8-dfac5dffd7fd) ## Highlights - **Unplanned Vacation With Smart Logistics** - Ian took a largely unplanned adults-only vacation to Scottsdale and intentionally relaxed by avoiding work at the pool. - He used Instacart/Costco deliveries to the resort to avoid expensive minibar prices and kept work light to reduce anxiety before leaving. Transcript: Ian Landsman Know what I love? Oh, no. You know what I love? I love when you take my advice. No. It's one of my favorite things in the whole world. When you say I'm right about something. No, no, no, no. I just love it. Aaron Francis I love it. Listen, we got to catch up. Regular listeners of the show will understand that Ian has never been right about anything. Ian Landsman That's true. Aaron Francis We have a long, we got a hundred something episodes of Ian just being constantly wrong. So wrong. I've been wrong once. And even then. Eh, eh. But for the new listeners, Ian's never been right. However, it brings me no joy. It brings me great pain, in fact, to say that vacation is awesome. Ian might have been right about taking a vacation. Vacations? If you haven't tried it, you gotta try it. It's pretty great. Ian Landsman Look at this guy going on vacation. Going on vacation. I knew you had to get away. You were on death week six or whatever. Hell March seven. Still sleeping in the office the week before you went on vacation. Aaron Francis Yep. Two hell two week. That's for sure. Ian Landsman We thought it was two hell two week eight. Seven, yeah. We just, yeah. So this seemed kind of last minute. Aaron Francis You're like, i'm not going to be here next week so give us a low down it was first minute for jennifer she planned the whole thing had it all sorted out kind of sneaked up on me it was last minute For me that's why i didn't warn you didn't talk about it on last week's episode. Didn't warn you because I just didn't really even know about it. I was at Bible study the week before and was telling people like, oh yeah, I'll be out of town next week. And Jennifer and I switch off weeks because babysitters and stuff. And they're like, where are you going? And I said, I'll be honest with you. I'm about 90% sure we're going to Scottsdale. That's all I know. Don't ask me where we're staying. What are we doing? I have no idea. Turns out, ([Time 0:00:20](https://share.snipd.com/snip/5c6ffb09-ea4a-49c9-bc86-9d665bf1c11e)) - **Slow Roll Public Launch To Avoid Moderation Overhead** - When launching a consumer platform, lean into a selective waitlist instead of opening broadly to avoid moderation headaches. - Ian plans to slow-roll outro.fm, curate early users, and remove undesired signups rather than build full-content moderation immediately. Transcript: Aaron Francis And I've talked about this a little bit before. Okay. Ian's conundrum corner. Here we are again. Ian Landsman So I've got a new website for Outdoor FM, which is kind of still just like a placeholder website. But, you know, it's part of the production app now. It's got a way you can add. When you sign up to be on the wait list, you can add your podcast feed. Okay. And so since I put this up, there's been some people who've signed up for the wait list. And now I can see their podcast when they sign up for the wait list. And some of them, you know, they're just, they're just bad things. You know, they're just bad. They're not even like, I'm not talking about left and right here. Okay. I'm not talking about left and right. I'm talking about bad things that most of the world is on board with them being bad. Oh no. And it's just like, you know, do I need the headache? Just give me a little bit of that. Like, do I need the headache? Like, are we talking like Nazi level? Yeah, we're talking like that kind of level. We're talking about like things we are. Now, listen. Now, is it real? Is it just somebody threw it in there, right? Like, I don't know. How do you already have Nazis? Right. But it's like, it just is like a little, a little, just a little bell in my head. They're like, do I need this hassle? Like you're in the content moderation game. It's like, yeah, like I'm in the B to C. Like, it's not, it's just like the way outro currently works is not even that much like public facing aspect to it. But, but first of all, there are features that would be more public facing that I would like to have. But even so, it's like when I'm doing support for this product and somebody's like, hey, like my Nazi podcast isn't working right. I'm trying to get this stuff going. I can't get my topics about Hitler lined up right and whatever. And it's like, man, like, do I want to be dealing with that? And like, I don't really want to be. But at the same time, do I know if that's real or not, right? Like, I don't know. Am I kicking people off the platform and going through that whole dog and pony show, you know, of like, I'm kicking you off. Like, I know, but there's a lot of headache there. You know what I'm saying? Aaron Francis No, there's no headache. This is merely a tool. You're not, you're not. I'm not preventing them from publishing. No, you're not Transistor. You're not Shopify. You're not cloudflare you're not de-platforming anyone you're just saying i don't want you as a customer easy case closed bento does that all the time bento does that all the time not With i'm sure with nazis but more with like fly by night uh shippers, brand new vibe coded products that have no human behind them. Jesse and his case is different because he has to protect IP reputations. But he's always just like, I don't like the company. Can't tell if there's a real human behind it. Looks like it was vibe slopped and sorry you're out um and it's the bar is just so much lower because you're not de-platforming like you're not saying you can't have your domain name you're Just saying use a different tool use trello use whatever um so i say just boot them say sorry not a fit for our platform don't say anything beyond that just say sorry not a fit see you later Ian Landsman I was also wondering with the idea too, of like, since this is not, this is not product A, this is not, I'm not depending on this thing. If I might just even, along the lines of what you're saying, maybe I just slow roll the whole thing along the lines of like, it's just waitlist for a long time. And I'm just picking and choosing, you know? I'm just picking and choosing and we're going to put our best foot forward with people I want to work with and kind of go from there, you know? Like, I don't need it to be open right away. And sometimes there's advantages to not being open right away. Like, people get into the kind of thing of like, how do I get into this? Aaron Francis If we can create a love to be part of an exclusive club. Right. Ian Landsman So there could be an element of that of like i mean i was gonna plan to kind of do that for a bit anyway because i didn't i was gonna say that fits that fits your historical behavior so far exactly We'll slow roll it yeah right exactly so i you know i want like me i want us and like two other people in there and then i want us and five other people in there. Like I was, I have no interest in like, I just open it and I deal with all the spam and all the crap. Right. And all that day one anyway. So like that was already kind of the plan, but I can kind of lean into that, get a feel for what's going on. Just see where I'm, see where I'm at with it., potentially. Yeah. But all right. You don't think I should be dissuaded? Aaron Francis No, don't be dissuaded. No. All right. I won't be dissuaded. If you have to have a no Nazis policy in your, there's a clause in your policies, that's fine. You could also just ghost them, leave them on the wait list forever. Ian Landsman I mean, the wait list, yeah, the wait list is easy, right? I'm just like, I already took them off the wait list, right? And that's fine. But yeah, you know, once if it's fully open and people are signed up, I'm not necessarily watching everybody who signs up. I don't have a bead on them every minute. So it's like now they're using it. And then you're like, oh, I'm kicking you off or whatever. It's a kind of annoying thing. I don't know. I can't imagine. Why don't I take my own advice, Aaron? B2B! Because you're never right. This is a good idea. Aaron Francis I can't imagine that with respect, I can't imagine that when you open the doors, you're going to be so inundated. Sworn with Nazis? Yeah, a simple Slack channel of who signed up is going to become so overwhelming that we can't see that the SS podcast signed up or whatever. What's the, there's a storm. There's like a... The Stormer, Daily Stormer. Yeah, that's what I was trying to think of. I'm not up on my Nazi culture. So if you see, like, DailyStormer.org or whatever it is, I don't think you're going to miss it between the, you know, thousand other signups. I think you should be okay. Ian Landsman I know. You know, there's just like, it's just headache. People, I don't want to deal with the headache. That's the thing. This is an excuse, Ian. Aaron Francis This is fear talking. It is fear talking. That's true. Okay. Ian Landsman All right. I will not be dissuaded for now we will not let ([Time 0:14:54](https://share.snipd.com/snip/aab06487-a41b-4625-90ae-b5b2ba83b6f1)) - **Raw Byte Diffs Fixed Terminal Serialization Bottleneck** - Solo's terminal performance hit a bottleneck from JSON serialization at 60 FPS, so Ian switched to a raw byte diff transport over channels. - Moving from JSON full-frames to byte-level patches and channels removed serialization overhead and enabled smooth 60 FPS terminal rendering. Transcript: Aaron Francis Ready so i will i have to make a mac os tahoe icon because Tahoe does weird things to the icon. They changed it. Right. I knew they changed that. Yeah. It's like you have to make it in layers so the glass can be applied programmatically, which is stupid. So I need to do that. And I have, I have a computer with Tahoe running so I can use the first-party icon designer thing, so it generates the right files. But that's it. That's the last thing I have to do. Wow. Okay. So the Hell Week previous, so before I left for vacation, like I said, I was working until three and then skipped church to my shame to keep working. It's just still, it was just still the terminal. The terminal, terminal, terminal. That is like, that is the whole app. Okay, so now the terminal is very rock solid. So scroll back. You can scroll back infinitely. There's a cap, but functionally infinitely. You can have as many processes running as you want, and you don't get any slowdown. You don't get any CPU, memory bloat anything like that um everything is like it's been it's been pushed to the extreme right and it now i haven't had i haven't had in i've been working with It all week and i haven't had a single a single issue now when i release it to the beta users within the first 14 seconds they will find an issue. I'm okay with that. Yeah. Because all of the ones that I know of are fixed. Yeah. So the final state of the terminal is thus. Everything is handled on the back end. That was the big unlock. WebGL on the front end, custom WebGL implementation on the front end. And then the most recent change, so this is like rewrite 4.7 or something. The most recent change is now over the wire, so from Rust to the web view that you have a couple of options one is the the tori ipc bridge that's a single bridge that we eliminated that a while Back so we use channels now which are like named named pipes where you can send multiple things to multiple different, well, it doesn't matter. The final upgrade, final, final V2 upgrade, is we're no longer using JSON over the wire. Ian Landsman Okay, that makes sense. Yeah. I didn't know you were, but I don't like the sound of that, so I think that this sounds better, whatever you're doing. Aaron Francis No, so the default is JSON. You just, like, if you send it through a channel, it like by default, JSON encodes it. And then you JSON decode it on the other side. That's normally fine. Unless you're trying to hit 60 frames a second on the terminal. You're making a bunch more data of just wrapper stuff. Then suddenly, for the first time in my career,son serialization and deserialization becomes a bottleneck and i'm like especially with bigger objects and stuff yeah i just relied On json to be this universal transport format and never once in my life thought about is this serialization going to become a bottleneck sufficient yeah so it became a bottleneck um And now and we only send patch sets so i don't send i'm not sending like whole frames i'm just sending this has changed but even that when you're sending you know i think we coalesced at 16 milliseconds so it maxes out at 60 frames a second over the wire but even then you send 60 messages over the wire in one second yeah and serializing and deserializing becomes a problem Um so we've now moved to a a no serialization uh transport and it's just i think it's called uh maybe flat message or something but it basically just, it's a different format. And it has like a few bytes at the beginning to do like some denotation of what's going on. And then we just send raw bytes over these channels. And the front end, in the back end, we don't encode it. In the front end, we don't decode it. It just, they both speak bytes. ([Time 0:36:20](https://share.snipd.com/snip/8fe0d63d-d279-4867-b329-49edec9b9e4d)) - **MCP Server Enables Autonomous Agent Orchestration** - The Solo MCP server acts as a man-in-the-middle enabling programs and agents to exchange input/output and timers, letting Solo drive processes autonomously. - Ian uses scratch pads and to-dos plus agent orchestration so agents create, update, lock, and close implementation tasks without losing context. Transcript: Aaron Francis Yeah. Are you ready for this? I love MCP servers. So one of the great realizations is that Solo, not Aaron, Solo is a man in the middle between your process and the way it displays, right? Yeah. So I don't ever see anything, just to be totally clear. But Solo does. And so Solo is the arbiter between the running process on the back end and like the input output and the front end, right? So this affords me a few things that I think are hard to come by. One of those is I can obviously do anything I want with all the output, right? Because it flows through me. Perhaps more interestingly, I can feed it input without user action, right? Like the program, solo the application can feed stuff in to your running processes without you having to take explicit keystroke action. So the MCP server has become central to, I think, what the core value of Solo is. Okay? Okay. So let me explain just a few things. We have two new primitives. So on your sidebar, you've got agents, commands, and terminals, right? There are two new primitives, which are totally opt-in. You don't have to use them if you don't want to. I'm excited about this. And that is scratch pads and to-dos. Ian Landsman Oh, yeah, that's right. You went to to-dos. I did. Aaron Francis So scratch pads just markdown documents. So I've got this beautiful markdown rendering pipeline. So just at the top here, I'll say I've been using scratch pads and to-dos like a madman. And I don't think I've written a single scratch pad or to-do. That's kind of like, that's kind of the unlock that I have discovered using solo is you get all these agents to do all this work. And then it's like, all right, well, I'm almost out of context. What do you want me to do now? And so I'll have them like research font rendering based on the current stack or whatever. And it's like, here are my 18 recommendations or things we need to look at. And then what do you do there? Do you say, all right, let's look at number one. And by the time it finishes looking at number one, it doesn't remember what the other ones are. It's lost track of everything. Do you write a markdown file onto disk? Maybe, maybe not, whatever. So now what I do is I say, okay, I want you to create a scratch pad out of that. And it just, it creates its own, it just knows what it is. And it writes the whole thing. And then I say, I'll either say like, all right, spawn a new agent, which is awesome. Spawn a new agent or spawn five new agents to read that scratchpad back, do further research and amend the scratchpad. Right. And so then it launches a clod, a codex and an amp and it goes off and it researches and then it writes it. It writes its findings to the scratchpad. So we're getting this better context, this better research. And then I'll point an agent at it and say, hey, can you read this scratchpad and create to-dos out of it? And then we've got these discrete little chunks that are trackable. And to-dos can have blockers. So I can say, great, write to-dos for everything that you see in the scratchpad. And then add links in the scratch pad to the to-dos so that we have this kind of reference going on. And then I can just, I feel confident that I've captured everything so I can close all my terminals, go home, play with the kids, right? And then I fire Solo back up, launch a fresh codex brand new codex that doesn't know me from adam and say hey can you read all the to-dos that are tagged with you know font rendering and start Start dispatching those to other clod and codex agents and it's like sure let me read all the to-dos oh great context uh all right i'll open a codex agent for that. And then I'll open a clod agent to review. And then, and it's just like all the context is totally there. The agents can lock to do's, they can comment on them, they can retag them, they can close them, they can update the body, whatever. ([Time 0:53:00](https://share.snipd.com/snip/a1df2674-5432-4a48-99b0-877bfbbf81dd)) - **Orchestrate Agents Instead Of Letting One Do Everything** - Use an orchestrator agent that never writes code but coordinates other agents to split work and prevent overlap. - Ian spawns Codex as an orchestrator that dispatches implementation agents, sets timers, and harvests results for QA with Claude. Transcript: Aaron Francis And it just, it creates its own, it just knows what it is. And it writes the whole thing. And then I say, I'll either say like, all right, spawn a new agent, which is awesome. Spawn a new agent or spawn five new agents to read that scratchpad back, do further research and amend the scratchpad. Right. And so then it launches a clod, a codex and an amp and it goes off and it researches and then it writes it. It writes its findings to the scratchpad. So we're getting this better context, this better research. And then I'll point an agent at it and say, hey, can you read this scratchpad and create to-dos out of it? And then we've got these discrete little chunks that are trackable. And to-dos can have blockers. So I can say, great, write to-dos for everything that you see in the scratchpad. And then add links in the scratch pad to the to-dos so that we have this kind of reference going on. And then I can just, I feel confident that I've captured everything so I can close all my terminals, go home, play with the kids, right? And then I fire Solo back up, launch a fresh codex brand new codex that doesn't know me from adam and say hey can you read all the to-dos that are tagged with you know font rendering and start Start dispatching those to other clod and codex agents and it's like sure let me read all the to-dos oh great context uh all right i'll open a codex agent for that. And then I'll open a clod agent to review. And then, and it's just like all the context is totally there. The agents can lock to do's, they can comment on them, they can retag them, they can close them, they can update the body, whatever. ([Time 0:55:54](https://share.snipd.com/snip/f52e86e8-37d6-42f9-831f-ee16d2a2047b)) - **Timers Replace Busy Waiting For Agent Workflows** - MCP timers let agents go idle and then notify orchestrators so they don't busy-wait, enabling long-running agent workflows. - Ian had agents set timers that fire when others go idle or after a set time, triggering new user-like messages into the orchestrator. Transcript: Aaron Francis And then using the solo MCP, this is great. You're going to love this. I want you to set a timer for yourself. So Solo MCP now exposes timers. I want you to set a timer for yourself. And the timer should fire when your implementation agents go idle. So when those implementation agents are done or they hit a permissions blocker, Solo will then notice, because, you know, the agent set a timer. Solo will notice that an agent went idle, and it will write a new user message to the orchestrator that says, heads up, this agent went idle. Here's the body of the timer that you requested me to send to you. And it just sends it as if it's a user message. And so now these orchestrator agents can be working for 10 or 15 minutes. The implementation agents can be working for 10 or 15 minutes. The orchestrator doesn't have to busy loop. It doesn't have to like set a bash sleep in the background. It doesn't just wait around. It just ends its turn, goes completely idle. And then solo will fire a timer when it has options for when any of the following agents go idle, when all of them go idle, or just a simple time. Like, after five minutes, check in on them. And so we now enter this loop where the orchestrator is, like, managing all of these bug-fixing agents, and it can write to the PTY so it can send input. It can also, of course, read output. So it'll wake up, it'll read like, okay, it said in all caps, no findings, which was what I told it to do. And so now that means it's done. I'm going to harvest all of what it did. And then I'm going to close that agent. And then I'm going to spawn a new one to do the QA because I always have Claude do the QA on it. And then it's like, we're now, we're now in agent orchestration mode. ([Time 0:58:16](https://share.snipd.com/snip/0c2bce1e-6571-4f78-be49-2c6d0d96e6e7)) - **Two Agents Debated Via MCP In Real Time** - Ian stress-tested MCP by spawning Claude and Codex with opposing viewpoints and watched them debate, exchange input, and set timers through the MCP. - The test confirmed agents could interact via MCP like users, including back-and-forth arguments and timer-driven messages. Transcript: Aaron Francis Um, and I warned each of it like, Hey, I want you to talk to this other agent. Codex, you're going to take the view of, uh, old earth, biblical creationism. Claude, you're going to take the view of young earth, biblical creationism. I want you to look at the MCP and I want you guys to start talking to each other. And I just watched these two agents debate each other, sending output, reading output, setting timers, like arguing. It was so surreal. Ian Landsman So ([Time 1:00:31](https://share.snipd.com/snip/63172b6a-aeaf-4ed4-bd5c-9c817d512989)) - **Keep Product Issues In PM Tools But Use Solo For Implementation** - Use Solo to manage implementation-level to-dos derived from product issues tracked in GitHub or Linear instead of replacing product trackers. - Ian pulls an issue into Solo, breaks it into implementation to-dos and has agents pick those up, then marks the original product issue done afterward. Transcript: Aaron Francis You pull that issue in into like working context. And then you got to figure out what are all the steps to fix this thing, right? That's where I think these solo to-dos live is all right i'm working on issue one two three let's do some research oh it looks like we need to touch these four areas we need to modify this Data structure great let's write to-dos organize them and then we can have other agents pick them up so it doesn't become it really is like your agent's to-dos and less like public facing Product to-dos. I happen to use it both ways because I am a solo developer. So I can go in and tell the agent like, hey, add a to-do because when I hit command E, it doesn't do this thing properly. That is a product facing to-do. But if I had a linear tracker, I'd say, pull this one, like, you know, leave a comment on this one and say, Aaron's working on it. Pull this in. Then internally, let's manage the actual implementation level to do's, which could be one, could be 10, depending on the feature. ([Time 1:05:16](https://share.snipd.com/snip/b36894e0-7d80-49de-85da-8148732eb2c8)) - **Process Visibility Prevents Agent Resource Leaks** - Solo shows agent spawn lineage, background process counts, and lets you kill subprocesses, which prevents runaway agent resource usage. - Ian added deep links, spawned-by metadata, and a footer that shows sub-process counts and PIDs to trace and kill agent processes quickly. Transcript: Aaron Francis It's full of agents right now. Yes. A few nice things. When an orca well when a when an agent spawns a new agent it uh automatically disables notifications so that like if it spawns 15 agents and then it sets a wake-up timer for itself i don't Need to hear this agent's done or like, hey, I finished my job. I don't want to hear that. So it automatically turns notification level to none, which is nice. And then in the bottom footer where there's meta information, it shows you which agent spawned the current one. So like if I go to an agent named, you know, to do 734 and I click on it at the bottom. I can see spawned by orchestrator and I can click on spawned by orchestrator and it jumps me to the orchestrator to it tab. So that's nice. A few other quality of life upgrades. Everything has a deep link. And so if I'm on a to do, I can right click on the to do and say copy link, and then go to the agent and say, look at this to do. So I don't have to say like, Hey, can you pull the to do that's titled whatever I just say? And it's, you know, solo colon slash slash project to do seven, whatever. Yeah, everything has a deep link. And then one other thing that I've added, which is quite nice, is you can see how many, you can see how many and what background processes an agent is running. And so, you know, sometimes you're like, why are my fans on? And then you look and it's like, oh, I've got literally 32 Vitesse runners running. So now in that little meta area down there, it'll show you like plus 17. And it's like, holy crap, I got 17 processes running? And you click on them and it shows you the process tree of like what's actually running. And you can kill them if you want. You can kill them straight from there. But there's also a copy all PIDs button. So you can copy all of the process IDs and then paste it into the agent and say, Hey, do you need any of these? Like, what the hell are you doing? ([Time 1:10:23](https://share.snipd.com/snip/75dfdcdf-ecc7-41b5-ba1c-d6cb537d3ba1)) - **Run A Tight Weekly AI Show For Clip Distribution** - Keep Token Town short and topical: pick one or two AI stories weekly and deliver tight 30-minute shows with clear takes. - Ian and Aaron plan a Thursday 2pm ET live stream then a polished podcast episode to stay focused and clip-friendly. Transcript: Ian Landsman This is why we need the other show where we're going to talk about that kind of stuff. But it all ties in here. Aaron Francis Tell us about the other show, Mr. Landsman. All right. Tokatown. Ian Landsman We talked to you about it two weeks ago. False start. I got sick. Aaron went on vacation. We were like, all right, we're're just gonna give it a couple weeks ai is not going anywhere it's gonna be here in a couple weeks it's two weeks later ai is still here so i think we're okay Um so this way we also changed the date we changed the reasons uh we didn't realize we did have a conflict with friends of the family uh the panel um but also we just came to our senses and Like Friday afternoon, that's not a good time to do what we're trying to do with this. So Thursdays, Thursdays, two o'clock Eastern, we are going to be live streaming token town. And then, uh, Friday, the episode will come out in podcast form and video form and all that stuff. So you keep an eye out on Twitter, on YouTube, 2 p.m. Eastern Thursday. We will be live, hopefully, if we don't get sick and nothing else weird happens, with the first episode. And we're going to talk about what's going on in AI, what you should be paying attention to, what we're doing, paying attention to. Yeah. And I'm excited to like start the process of feeling it out, what the show it wants to be. You know, I think we're going to start with a premise and then we're going to see where it goes. And we're going to see how the live stream impacts it. We've never really live streamed on this show. So that's going to add a different dynamic, how we integrate live feedback, if we integrate live feedback, what we want to do with all that. But, yeah, so we're going to put our toe in the water there and see how it goes. But I'm excited. I'm ready. It's going to be great. Thursday, 2 p.m. Aaron Francis Thursday, 1 p.m. Central. Yes. That's the best time zone. That is the worst time zone. Of the show, for those of you that are new to Token Town, Token Town is the tight what's happening in AI, hot takes, news, information. If you want the how's your father, you stay here on Mostly Technical. If you want tight AI, you also stay here, but then you add on Token Town. That's right. Ian Landsman You don't leave here. Nobody should leave here. Nobody leaves. Nobody leaves. This is additive only. We're not going to do a half hour on solo like we just did on Token Town. This is for the half hour on how's it going solo, how's it going with outro, how's it going with help spot, how's it going with faster. That is here. That's not going to be over there, which is going to be hard for us. I'm interested to see. I know. Because we are going to be inclined to be like, hey, yesterday I was working on this thing that's totally relevant to this. Wrong show. And we need to do some of that, but we need to, it needs to just be tight. It's like a mention. It's like, yeah, we have experience with this, but it can't be like what we did with it. So, I don't know that's all interesting stuff navigate we're trying to keep that show to 30-ish minutes just to kind of keep it tight but i don't know if that's even realistic with i mean I feel like we can read we can read just straight read from a list everything that's happened in the past week and that would take more than half an hour so i agree there's going to be some Editing required which is the point of the show. Like, yes, 32 things happened on yesterday. Like, we're going to pick out the one or two we think is kind of important. And then, yeah, come up with our takes. Try to keep it tight. We'll see how that goes. Aaron Francis We'll see how that goes. Regardless, it's going to be good. That much I know. It's going to be good. It's going to be good. Ian Landsman It's is highly important. That's the main, it needs to be informative and entertaining. And we're all in this together with the AI stuff and we're all navigating. So that's the goal of the show. And we will see the exact path of how we get there. We're going to experiment all together. Aaron Francis Should we announce the headline sponsor just to give them a little bonus here? The headline sponsor for Token Town, tokentown.com, is a new sponsor named Bento. Maybe you've heard of them. Hey, Bento. We love Bento. You'll hear more about what Bento actually is because we've never talked about them here before. But they are sponsoring Token Town as the headline sponsor. So thanks to Jesse for making this this fever dream possible. Ian Landsman Just great sponsors you have a good relationship with. It's like, I just went to Jesse. I was like, hey, what do you think about this new thing? And he's like, yes, I'm in. And we pushed it about two weeks and he's fine. And we're all fine. And like, it's going to be great. It's great to have his support. Thank you so much, Jesse, for that. Yeah. And if anybody else out there interested in sponsoring the new show, it's going to be obviously, I think early will be kind of this audience, of course. ([Time 1:16:09](https://share.snipd.com/snip/b9d22d70-419b-4dac-a72d-fccdb2c8ad5d))